range operation of outer cache when start >= end?

range operation of outer cache when start >= end?

[Masahiro Yamada]

Hi.


I know I am nitpicking.  Forgive me if I am asking a silly question.


How should the outer-cache handle such an insane case like  start >= end?

Assumed answers are:

[1] Do not care about that.  It should never happen.  If it does, fix
the caller.

[2]  Add   "if (start < end)" checking to
      outer_inv_range(), outer_clean_range(), outer_flush_range()
      in arch/arm/include/asm/outercache.h

[3] It should be cared in each of callbacks as needed



For example, l2c210_inv_range()
flushes the both ends of the range even if start >= end.

In this case, there is nothing to do, so should it have something like this?
    if (start >= end)
                 return;


static void l2c210_inv_range(unsigned long start, unsigned long end)
{
        void __iomem *base = l2x0_base;

        if (start & (CACHE_LINE_SIZE - 1)) {
                  start &= ~(CACHE_LINE_SIZE - 1);
                  writel_relaxed(start, base + L2X0_CLEAN_INV_LINE_PA);
                  start += CACHE_LINE_SIZE;
        }

        if (end & (CACHE_LINE_SIZE - 1)) {
                  end &= ~(CACHE_LINE_SIZE - 1);
                  writel_relaxed(end, base + L2X0_CLEAN_INV_LINE_PA);
        }

        __l2c210_op_pa_range(base + L2X0_INV_LINE_PA, start, end);
        __l2c210_cache_sync(base);
}





-- 
Best Regards
Masahiro Yamada

Re: range operation of outer cache when start >= end?

[Russell King - ARM Linux]

On Fri, Jan 08, 2016 at 11:54:30AM +0900, Masahiro Yamada wrote:
> I know I am nitpicking.  Forgive me if I am asking a silly question.
> 
> 
> How should the outer-cache handle such an insane case like  start >= end?

Passing start >= end isn't defined, code should not pass start >= end.

> Assumed answers are:
> 
> [1] Do not care about that.  It should never happen.  If it does, fix
> the caller.

This applies.  What situation are you seeing start >= end?

What you will get with the existing code is potentially some cache
cleaning and a sync, but nothing apart from that.  __l2c210_op_pa_range()
becomes a no-op of start >= end.  However, that behaviour is not
guaranteed.

-- 
RMK's Patch system: http://www.arm.linux.org.uk/developer/patches/
FTTC broadband for 0.8mile line: currently at 9.6Mbps down 400kbps up
according to speedtest.net.

Re: range operation of outer cache when start >= end?

[Masahiro Yamada]

Hi Russell,


2016-01-09 0:06 GMT+09:00 Russell King - ARM Linux <linux@arm.linux.org.uk>:
> On Fri, Jan 08, 2016 at 11:54:30AM +0900, Masahiro Yamada wrote:
>> I know I am nitpicking.  Forgive me if I am asking a silly question.
>>
>>
>> How should the outer-cache handle such an insane case like  start >= end?
>
> Passing start >= end isn't defined, code should not pass start >= end.
>
>> Assumed answers are:
>>
>> [1] Do not care about that.  It should never happen.  If it does, fix
>> the caller.
>
> This applies.  What situation are you seeing start >= end?

I never see such a case at all.
When I saw l2c210_inv_range(), I just wondered whether it should be
cared or not.


> What you will get with the existing code is potentially some cache
> cleaning and a sync, but nothing apart from that.  __l2c210_op_pa_range()
> becomes a no-op of start >= end.  However, that behaviour is not
> guaranteed.

Thanks for your explanation!

-- 
Best Regards
Masahiro Yamada